Research map / Architectures and efficiency / RNNs and normalization
Deep Networks with Stochastic Depth
Gao Huang, Yu Sun, Zhuang Liu, Daniel Sedra, Kilian Q. Weinberger
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add stochastic depth | ResNet-110 · CIFAR-10+ | test error (%): 6.41 -> 5.25 | Improved |
| add stochastic depth | ResNet-110 · CIFAR-100+ | test error (%): 27.22 -> 24.98 | Improved |
| add stochastic depth | ResNet-110 · CIFAR-10 (no augmentation) | test error (%): 13.63 -> 11.66 | Improved |
| add stochastic depth | ResNet-110 · CIFAR-100 (no augmentation) | test error (%): 44.74 -> 37.8 | Improved |
| add stochastic depth | ResNet-152 · SVHN | test error (%): 1.80 -> 1.75 | Improved |
| add stochastic depth | ResNet-1202 · CIFAR-10 | test error (%): 6.67 -> 4.91 | Improved |
| add stochastic depth, 90 epochs | ResNet-152 · ImageNet val | top-1 error (%): 23.06 -> 23.38 | Hurt |
| add stochastic depth with 120 epochs vs baseline 120 epochs | ResNet-152 · ImageNet val | top-1 error (%): 21.78 -> 21.98 | No clear effect |
| increase depth from 110 to 1202 layers with stochastic depth (ablation) | ResNet · CIFAR-10 | test error (%): 5.25 -> 4.91 | Improved |
| increase depth from 110 to 1202 layers without stochastic depth (ablation) | ResNet constant depth · CIFAR-10 | test error (%): 6.41 -> 6.67 | Hurt |
| use linear decay instead of uniform survival probability (ablation) | stochastic depth · CIFAR-10 | test error (%) | Improved |
| stochastic depth with p_L=0.2 (ablation) | stochastic depth linear decay · CIFAR-10 | training time: 40% reduction in training time | Improved |
Closest papers by meaning
- Deep Layers as Stochastic Solvers (ICLR 2019 · similarity 0.89)
- Regularization in ResNet with Stochastic Depth (NeurIPS 2021 · similarity 0.88)
- A Mean Field Analysis Of Deep ResNet And Beyond: Towards Provably Optimization Via Overparameterization From Depth (ICML 2020 · similarity 0.88)
- A Convergence Theory for Deep Learning via Over-Parameterization (ICML 2019 · similarity 0.87)
- Tensor Programs VI: Feature Learning in Infinite Depth Neural Networks (ICLR 2024 · similarity 0.86)
- Stochastic Kernel Regularisation Improves Generalisation in Deep Kernel Machines (NeurIPS 2024 · similarity 0.86)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
